[🐞 BUG] Error occurred while enabling the model

[RussQuan]

Describe the bug

When I try to enable the model, I receive the following error:

Failed to build the model execution plan using a model architecture file '/Users/pi/Library/Caches/FluidAudio/CompiledNemotronModels/encoder-842ccec78631ca4e/encoder_ane_1781622167444.mlmodelc/model.mil' with error code: -14.

To Reproduce

Steps to reproduce the behavior:

1. Open FluidAudio.
2. Navigate to the Models settings page.
3. Download and enable the Nemotron model.
4. Wait for the model initialization process.
5. An error message appears and the model cannot be enabled.

Expected behavior

The model should be enabled successfully and become available for use.

Screenshots

Environment

• macOS Version: 15.7.3 (24G419) Macbook air m4 32G
• App Version: FluidVoice v1.5.15
• Architecture: Apple Silicon

Additional context

I am using a MacBook Air M4. The issue occurs immediately when attempting to enable the Nemotron model after downloading it. Re-downloading the model and restarting the application did not resolve the issue.

Crash Logs

The application did not crash, but the following error was shown:

Failed to build the model execution plan using a model architecture file '/Users/pi/Library/Caches/FluidAudio/CompiledNemotronModels/encoder-842ccec78631ca4e/encoder_ane_1781622167444.mlmodelc/model.mil' with error code: -14.

[altic-dev]

Open terminal and run

rm -rf /Users/pi/Library/Caches/FluidAudio/CompiledNemotronModels

Wanna delete this and see if it works?

[mvanhorn]

Implemented in #456 - clears the corrupt FluidAudio CompiledNemotronModels cache before the GPU fallback retry so CoreML recompiles fresh, with an actionable error if it still fails.